Skip to content

Fully supervised binary classification of skin lesions from dermatoscopic images using multi-color space moments/texture features and Support Vector Machines/Random Forests.

Notifications You must be signed in to change notification settings

anindox8/Multi-Color-Space-Features-for-Dermatoscopy-Classification

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

40 Commits
 
 
 
 
 
 
 
 

Repository files navigation

Color/Texture Features with Support Vector Machine and Random Forest for Disease Classification of Skin Lesions

Problem Statement: Fully supervised binary classification of skin lesions from dermatoscopic images.

Note: The following approach won 1st place in the 2019 Computer-Aided Diagnosis: Machine Learning in Dermascopy Challenge at Universitat de Girona scoring 88.1% accuracy (kappa: 0.732) at test-time, during the 2018-20 Joint Master of Science in Medical Imaging and Applications (MaIA) program.

Acknowledgments: Mina Sami for the Python implementation of Shades of Gray Color Constancy.

Data: Class A: Nevus; Class B: Other (Melanoma, Dermatofibroma, Pigmented Bowen's, Basal Cell Carcinoma, Vascular, Pigmented Benign Keratoses) [4800/1200/1000 : Train/Val/Test Ratio]

Directories
● Data I/O Functions: scripts/dataio.py
● Preprocessing Functions: scripts/preprocess.py
● Unsupervised Segmentation Functions: scripts/segment.py
● Feature Computation Functions: scripts/colorfeatures.py
● Final Feature Extraction Function: scripts/feature_extraction.py
● Classifier Support Functions: scripts/classify.py
● Inference Pipeline Notebook: scripts/predict.ipynb
● Training-Validation Pipeline Notebook: scripts/train-val.ipynb

Color Constancy

Color Constancy

Occlusion Removal

Hair Removal

Unsupervised Segmentation

Unsupervised Segmentation

Color Space

Color Space

Gabor Filter Features

Gabor Filter Features

HOG Features

HOG Features

Feature Selection

Feature Selection

Experimental Results

Experimental Results

About

Fully supervised binary classification of skin lesions from dermatoscopic images using multi-color space moments/texture features and Support Vector Machines/Random Forests.

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published